
To benchmark shared hosting fairly, put the same test site on each host, measure speed from the locations where your visitors are, monitor uptime for at least a month, and compare the total cost over several years including renewals. A single speed test on a single day tells you very little. This guide sets out a method you can repeat yourself, so your decision rests on your own measurements rather than someone else’s league table.
Why most hosting benchmarks mislead
Published comparisons often test each host once, from one location, with different sites, at different times, and compare introductory prices with no mention of renewals. Results then reflect the test, not the host. Shared hosting performance also varies with what other accounts on the same server are doing, so short tests can catch a host on a good or bad day. A useful benchmark controls for these variables.
Step 1: define what you are testing
Decide which plans you are comparing and why. Compare plans at a similar price and with similar resources, and write down for each: storage, storage type, bandwidth, number of sites, databases and mailboxes, whether SSL and backups are included, the web server, PHP versions available, and the server location you were given. Note the date, because plans change.
Step 2: build one identical test site
- Use the same platform you will run in production, for example WordPress with the same theme and a realistic amount of content and images.
- Install the same plugins on every host, and decide in advance whether you will test with caching on, off or both. Test both if you can: uncached results show raw server speed; cached results show what visitors will usually experience.
- Use the same PHP version everywhere.
- Point a separate subdomain at each host so DNS is comparable.
Step 3: measure speed properly
What to measure
- Time to First Byte (TTFB): the delay before the server starts sending the page. It is the measure hosting affects most directly.
- Largest Contentful Paint (LCP): when the main content appears. Google rates 2.5 seconds or less as good.
- Full page load as a secondary figure.
How to measure
- Test from at least two locations, including the one closest to your visitors.
- Run tests at different times of day over at least a week, and report the median rather than the best result.
- Use the same tool and settings for every host, such as WebPageTest or PageSpeed Insights, and keep the raw results.
- Warm the cache with a first request before measuring cached performance.
Step 4: test under load, with permission
A site that is fast for one visitor may slow sharply when many arrive at once. Load-testing tools such as k6 or Apache Bench can simulate concurrent visitors and show where response times start to climb or errors appear. On shared hosting, heavy load tests can breach the host’s acceptable use policy and affect other customers, so ask the host first, keep tests short and modest, and stop at the first errors.
Step 5: monitor uptime for at least 30 days
Use an external monitoring service that checks the site every minute from more than one location, and record every outage with its start time and duration. Thirty days is a minimum; three months is better. Then compare the result with each host’s written commitment. An uptime figure in a service level agreement with credits means more than one on a sales page. For reference, 99.9% availability allows about 43 minutes of downtime in a 30-day month.
Step 6: calculate the real cost
Compare the total cost over three years, not the first-month price:
- First-term price for the term you would actually buy.
- Renewal price for the remaining years.
- Add-ons you would need: SSL, backups, email, a domain, malware scanning.
- Any migration or setup fees.
Hosts with introductory pricing can look cheap in month one and expensive by year three. Tremhost, for example, has no introductory pricing: its web hosting renews at the sign-up price, from $30 a year on the Himalaya plan, and includes free SSL, unlimited mailboxes and daily backups.
Step 7: test support and recovery
- Send each host the same technical question and record how long a useful answer takes, through each channel offered.
- Restore a file and a database from the host’s backups and time it.
- Check how PHP versions, cron jobs and email are managed in the control panel.
Step 8: score and publish honestly
Decide the weighting before you look at results, for example speed 30%, uptime 30%, three-year cost 25% and support 15%. Report medians, the test dates, locations, plans and anything that went wrong. If you publish your findings, include the raw data and say who paid for the hosting. Readers trust results they can check.
Frequently asked questions
How long should a hosting benchmark run?
At least 30 days for uptime and a week of repeated speed tests at different times. Longer is better.
Should I test with caching on or off?
Both, if possible. Uncached results show server capability; cached results reflect what most visitors see.
Can I trust a single PageSpeed score?
No. Scores vary between runs. Use the median of many runs and look at TTFB and LCP rather than the overall score alone.
Next step
If you want to include Tremhost in your test, compare the plans on the Tremhost web hosting page; there is a 30-day refund on your first web hosting order.
